Currently, when searching for a to-be-searched-for target object by using a network, generally, a user inputs or selects some known search criteria for the target object, and then searches for the target object according to the search criteria. For example, searching may be performed according to a price range of the target object, or searching may be performed according to an area to which the target object belongs.
In the foregoing object search solution, a user is required to be able to clearly describe a search criterion on which a search is based. For example, the search criterion may be selected from some preset criteria, or may be directly input. However, in an actual application, when searching for a target object, a user may be unable to clearly describe a search criterion expected by the user. For example, the user expects to search for an object of a particular color, but the color cannot be accurately described using a name of a known color, or the user expects to search for an object of a particular shape, but the shape is not regular. Consequently, in this case, the user cannot search for the target object expected by the user. In addition, even if searching is performed, it is possible that a search result does not meet a search intention of the user because the search criterion is not accurate, thereby leading to a relatively poor search effect. An object search method in the prior art cannot provide the user with a more flexible search manner, and is relatively greatly restricted by an application scenario.